objc-parse.y and bison 1.875

Joseph S. Myers jsm28@cam.ac.uk
Mon Jan 13 05:17:00 GMT 2003


On Thu, 9 Jan 2003, Mike Stump wrote:

> objc-parse.y doesn't build anymore with bison 1.875.  Bison now expects 
> 0 reduce/reduce conflicts and objc-parse.y has 1 such conflict, causing 
> a hard error from bison.  In the past, this was non-fatal in bison.
> 
> We need to either mandate an old bison, or fix the reduce/reduce or get 
> bison to allow us to expect 1 reduce/reduce conflict or...

Getting rid of the reduce/reduce conflict is correct.

There have been suggestions of merging cc1obj with cc1 (and some changes
to a runtime flag_objc in this direction, replacing objc as a separate
C-family language in shared code).  If these lead to a single parser based
on the current parser but with runtime rather than compile-time
conditionals, I'll disapprove them unless the reduce/reduce conflict is
removed and there's a proper analysis of all the extra shift/reduce
conflicts for objc and why they don't affect C.

-- 
Joseph S. Myers
jsm28@cam.ac.uk



More information about the Gcc-bugs mailing list